You asked for comment.
I believe the cut-off level should remain at the 99.9999 %ile.
The Society is one of Dr. Hoeflin's achievements and has nothing to do with me either.
If there is a test problem with available tops then this can be solved simply by administering one of the two currently used tests and then retesting all applicants meeting a set point where few will be missed by a more advanced test. If the error of measurement is 1/3 sigma retesting all those 2/3 sigma below the ceilings of the existing test will lose only 25% of those who would qualify on the higher test. If you move the cut-off to 1 sigma below the ceiling you will lose 10% of them. If you move it to 1-1/3 sigma below the cut-off you lose 2%. So my suggestion is that this figure be the accepted one since it is unlikely even the superior test will reach such a comparable level, i.e., it is never going to be likely that it is 98% accurate in picking the correct candidate.
The final test need only to be a quite short one--say 20 to 30 items. The difficulty would be set so that 50% of items scored correct would reach the 99.9999 %ile. The difficulty level can be placed above the level of the final people--thus almost all candidates would go bust doing it.
A further requirement would be that the multiple choices given would leave them all with the impression they'd won through. This would prevent answers being given out and/or test copies floating around. For this to be possible all the answers given would need to appear very plausible.
Another way of doing this is to increase the sheer number of choices given to testees--say to 20 choices.
Dr. Hoeflin, you, and I have by now enough experience with test construction to put a final test together that will meet all the above requirements.
I have been toying with the idea of a mindstretching test that will call upon the span of the person's ability to keep the concept constant that can have its items generated by a computer, e.g., writing a computer program than can generate larger and larger sets of involved rules thus allowing almost any number of such items to be spawned, the only effort being in the initial construction of the program--but that is a single concept.
You may choose to reject such a notional idea.
Humanity in the collective produces only the odd great man or two in any century. We must see ourselves in light of this. Your work in test construction has been championed by two of psychology's luminaries and doing a straw poll of our Mega membership would, I'm sure, produce a few more comparables and many big achievers in general. So we are already out of proportion to our numbers.